Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / volcengine/verl issues and pull requests
#134 - DRPO KL term position
Issue -
State: open - Opened by jcao-ai 7 days ago
#133 - [Liger-kernel] Add an option to use `AutoLigerKernelForCausalLM` to load model
Pull Request -
State: open - Opened by hongpeng-guo 7 days ago
#132 - [SFT] Support context parallelism for SFT
Pull Request -
State: open - Opened by xingyaoww 8 days ago
- 3 comments
#131 - 微信交流群满了
Issue -
State: closed - Opened by Unakar 8 days ago
- 1 comment
#130 - Update README.md
Pull Request -
State: closed - Opened by vermouth1992 8 days ago
#129 - [ppo] refactor: refactor old_log_prob into a separate function
Pull Request -
State: closed - Opened by vermouth1992 8 days ago
#128 - [ci] feat: add ci for sft trainer
Pull Request -
State: closed - Opened by PeterSH6 9 days ago
#127 - [SFT] feat: Add LoRA support for SFT
Pull Request -
State: closed - Opened by xingyaoww 9 days ago
- 4 comments
#126 - vllm-npu-support
Pull Request -
State: closed - Opened by Chendong98 9 days ago
#125 - [misc] fix: normalize batch size should divide sp size
Pull Request -
State: closed - Opened by PeterSH6 10 days ago
#124 - [algo] feat: support GRPO algorithm
Pull Request -
State: closed - Opened by PeterSH6 10 days ago
#123 - [perf] feat: support meta device init and parallel load for fsdp
Pull Request -
State: closed - Opened by zhiqi-0 10 days ago
#122 - [perf] enable multiproc dataloader in sft trainer
Pull Request -
State: closed - Opened by hiyouga 11 days ago
#121 - [perf] feat: support ref/rm offload
Pull Request -
State: closed - Opened by vermouth1992 11 days ago
#120 - [misc] fix: super tiny fix mlflow error
Pull Request -
State: closed - Opened by fzyzcjy 12 days ago
#119 - Invalid value "timing(s)/gen" for parameter 'metrics[39].name' supplied: Names may only contain alphanumerics, underscores (_), dashes (-), periods (.), spaces ( ) and slashes (/).
Issue -
State: closed - Opened by fzyzcjy 12 days ago
- 2 comments
#118 - [perf] feat: Support dynamic batch size
Pull Request -
State: closed - Opened by vermouth1992 13 days ago
#117 - [misc] feat: support mfu calculation
Pull Request -
State: closed - Opened by vermouth1992 14 days ago
#116 - [test] Add tests for SPMD vLLM
Pull Request -
State: open - Opened by ZSL98 15 days ago
#115 - [dataproto] fix: add assertion for uneven chunk
Pull Request -
State: closed - Opened by vermouth1992 15 days ago
#114 - [perf] fix: set use_reentrant=False when enable gradient checkpointing
Pull Request -
State: closed - Opened by vermouth1992 15 days ago
#113 - [ci] fix: change VLLM_ATTENTION_BACKEND to XFORMERS to avoid illegal memory access
Pull Request -
State: closed - Opened by vermouth1992 15 days ago
#112 - [ci] fix: add force stop in ray e2e ci to clean env
Pull Request -
State: closed - Opened by PeterSH6 15 days ago
#111 - [misc] chore: refactor and add several metrics
Pull Request -
State: closed - Opened by vermouth1992 16 days ago
#110 - [misc] fix: fix license
Pull Request -
State: closed - Opened by vermouth1992 16 days ago
#109 - [misc][Long Context] feat: support ulysses for long context training
Pull Request -
State: closed - Opened by PeterSH6 16 days ago
- 7 comments
#108 - nccl error when using multi node training
Issue -
State: closed - Opened by Cppowboy 17 days ago
- 2 comments
#107 - [readme] docs: add acknowledgement
Pull Request -
State: closed - Opened by eric-haibin-lin 17 days ago
- 1 comment
#106 - feature request: support different versions of vllm
Issue -
State: open - Opened by Cppowboy 17 days ago
- 5 comments
#105 - [misc] feat: add Ray Summit Youtube video link
Pull Request -
State: closed - Opened by vermouth1992 18 days ago
#104 - Questions about collocating difference roles
Issue -
State: closed - Opened by hashword0428 19 days ago
- 1 comment
#103 - [BREAKING][refactor] feat: hybrid_engine dir to sharding_manager for more general repres…
Pull Request -
State: closed - Opened by PeterSH6 19 days ago
#102 - Fix the displayed loss in the sft trainer for gradient accumulation > 1
Pull Request -
State: closed - Opened by hiyouga 19 days ago
#101 - Support saving to huggingface
Issue -
State: closed - Opened by rawsh 20 days ago
- 2 comments
#100 - [misc] feat: support different flash_attn versions with variable num returns
Pull Request -
State: closed - Opened by PeterSH6 20 days ago
#99 - [misc] fix reward model issue with TokenClassification model and support running particular steps instead of epochs
Pull Request -
State: closed - Opened by PeterSH6 20 days ago
#98 - [doc] fix readme link issue and add citation
Pull Request -
State: closed - Opened by PeterSH6 21 days ago
#97 - Fused CE loss integration
Issue -
State: open - Opened by eric-haibin-lin 21 days ago
- 1 comment
Labels: help wanted
#96 - Liger kernel integration
Issue -
State: open - Opened by eric-haibin-lin 21 days ago
- 4 comments
Labels: help wanted
#95 - Several issues on current main
Issue -
State: closed - Opened by vermouth1992 21 days ago
- 1 comment
#94 - [example] fix: fix notebook link due to username update
Pull Request -
State: closed - Opened by eric-haibin-lin 21 days ago
#93 - [FSDP] optimizer offload
Issue -
State: closed - Opened by eric-haibin-lin 22 days ago
- 1 comment
Labels: fsdp
#92 - [example] docs: add getting started notebook with free GPUs from lightning
Pull Request -
State: closed - Opened by eric-haibin-lin 22 days ago
#91 - [misc] feat: spport rmpad/data-packing in FSDP with transformers
Pull Request -
State: closed - Opened by PeterSH6 23 days ago
- 21 comments
#90 - [misc] fix: fix validation dp_size. fix #78
Pull Request -
State: closed - Opened by vermouth1992 24 days ago
#89 - [megatron] docs: clean up unused code, update megatron backend docs and installation docs
Pull Request -
State: closed - Opened by eric-haibin-lin 24 days ago
#88 - [docker] megatron: add TE to ngc dockerfile
Pull Request -
State: closed - Opened by eric-haibin-lin 24 days ago
- 2 comments
#87 - Examples for slurm/multi-node ppo training setup
Issue -
State: closed - Opened by awcvec 24 days ago
- 8 comments
#86 - [misc] chore: remove useless files
Pull Request -
State: closed - Opened by vermouth1992 24 days ago
#85 - Support megatron 0.6 in veRL
Pull Request -
State: open - Opened by Chendong98 25 days ago
- 2 comments
#84 - add changes from prime codebase
Pull Request -
State: closed - Opened by xingyaoww 26 days ago
#83 - [misc] feat: add several useful functions in protocol
Pull Request -
State: closed - Opened by PeterSH6 27 days ago
- 2 comments
#82 - No module named 'megatron.optimizer'
Issue -
State: closed - Opened by Wraythh 27 days ago
- 1 comment
#81 - Fix repeat
Pull Request -
State: closed - Opened by caoshiyi 27 days ago
- 1 comment
#80 - [rollout] feat: support best-of-n generation in vLLM
Pull Request -
State: closed - Opened by PeterSH6 27 days ago
#79 - [Pending SGLang] Support torch.compile and allow disabling FSDP
Pull Request -
State: open - Opened by fzyzcjy 28 days ago
#78 - Validation dataset silently drops last batch
Issue -
State: closed - Opened by fzyzcjy 28 days ago
- 3 comments
#77 - RayTaskError
Issue -
State: closed - Opened by Raf-Chen 28 days ago
- 2 comments
Labels: bug, vllm related
#76 - Resume from checkpoints
Issue -
State: open - Opened by fzyzcjy 29 days ago
- 2 comments
#75 - Tiny refactor and cleanup
Pull Request -
State: open - Opened by fzyzcjy about 1 month ago
- 2 comments
#74 - Support Mlflow, allow forward to have different batch size, compute more metrics
Pull Request -
State: closed - Opened by fzyzcjy about 1 month ago
- 1 comment
#73 - Question about actor training-rollout resharding
Issue -
State: open - Opened by 0oshowero0 about 1 month ago
- 2 comments
#72 - [community] docs: fix WeChat link
Pull Request -
State: closed - Opened by eric-haibin-lin about 1 month ago
#72 - [community] docs: fix WeChat link
Pull Request -
State: closed - Opened by eric-haibin-lin about 1 month ago
#71 - [community] docs: fix WeChat link
Pull Request -
State: closed - Opened by eric-haibin-lin about 1 month ago
#71 - [community] docs: fix WeChat link
Pull Request -
State: closed - Opened by eric-haibin-lin about 1 month ago
#70 - [doc] fix: experiment section url
Pull Request -
State: closed - Opened by PeterSH6 about 1 month ago
#70 - [doc] fix: experiment section url
Pull Request -
State: closed - Opened by PeterSH6 about 1 month ago
#69 - does this framework support long-generation such 8k-16k
Issue -
State: open - Opened by yyht about 1 month ago
- 2 comments
#69 - does this framework support long-generation such 8k-16k
Issue -
State: open - Opened by yyht about 1 month ago
- 1 comment
#68 - Support RLOO/GRPO/REINFORCE?
Issue -
State: open - Opened by fzyzcjy about 1 month ago
- 24 comments
#68 - Support RLOO/GRPO/REINFORCE?
Issue -
State: open - Opened by fzyzcjy about 1 month ago
- 24 comments
#67 - Super tiny fix typo in title
Pull Request -
State: closed - Opened by fzyzcjy about 1 month ago
- 2 comments
#67 - Super tiny fix typo in title
Pull Request -
State: closed - Opened by fzyzcjy about 1 month ago
- 2 comments
#66 - Missing doc about "Algorithm Baselines"
Issue -
State: closed - Opened by fzyzcjy about 1 month ago
- 6 comments
#65 - [misc] fix: weak reference of WorkerDict in RayTrainer
Pull Request -
State: closed - Opened by PeterSH6 about 1 month ago
#65 - [misc] fix: weak reference of WorkerDict in RayTrainer
Pull Request -
State: closed - Opened by PeterSH6 about 1 month ago
#64 - Actor model didn't update correctly when upgrade megatron to core-r0.6.0
Issue -
State: open - Opened by Wodswos about 1 month ago
- 7 comments
#63 - [install] chore: add pyproject.toml. make vllm default dependency
Pull Request -
State: closed - Opened by eric-haibin-lin about 1 month ago
#62 - package confilct
Issue -
State: closed - Opened by hljjjmssyh about 1 month ago
- 1 comment
#61 - [misc] feat: remove @ray.remote on workers to allow inheritance
Pull Request -
State: closed - Opened by PeterSH6 about 1 month ago
#60 - fix response_mask index
Pull Request -
State: closed - Opened by huiyeruzhou about 1 month ago
- 2 comments
#59 - [install] fix: revert pyproj.toml and fix tensordict req
Pull Request -
State: closed - Opened by eric-haibin-lin about 1 month ago
#58 - FIRE sampling added.
Pull Request -
State: open - Opened by laonahongchen about 1 month ago
#57 - Add pyproject.toml
Pull Request -
State: closed - Opened by pcmoritz about 1 month ago
- 1 comment
#56 - [example] docs: improve the quickstart documentation
Pull Request -
State: closed - Opened by eric-haibin-lin about 2 months ago
#55 - [ppo] chore: remove unused flash_attn dependency, and add docs for GSM8k reward
Pull Request -
State: closed - Opened by eric-haibin-lin about 2 months ago
- 1 comment
#54 - [algorithm] docs: add steps to reproduce PPO algorithm results
Pull Request -
State: closed - Opened by eric-haibin-lin about 2 months ago
#53 - Do we have plans for data packing?
Issue -
State: closed - Opened by YixinSong-e about 2 months ago
- 7 comments
#52 - [BREAKING][refact]: move actor/critic/hybrid_engine/reward_model/rollout/workers out of ppo directory for reuse
Pull Request -
State: closed - Opened by PeterSH6 about 2 months ago
#51 - Question about recomputation in actor module
Issue -
State: closed - Opened by 0oshowero0 about 2 months ago
- 3 comments
#50 - (fix): fix values response mask in dp critic.
Pull Request -
State: closed - Opened by PanAndy about 2 months ago
- 4 comments
#49 - [sft] feat: fix sft dataset with latest preprocess code
Pull Request -
State: closed - Opened by eric-haibin-lin about 2 months ago
#48 - Calling for Improving Robustness of FSDP-vLLM Rollout
Issue -
State: closed - Opened by nwiad about 2 months ago
- 4 comments
#47 - api: rename tracking logger to wandb logger type
Pull Request -
State: closed - Opened by eric-haibin-lin about 2 months ago
#46 - [ray] latest ray compatibility
Issue -
State: closed - Opened by eric-haibin-lin about 2 months ago
- 1 comment
#45 - [BREAKING][core] move single_controller into verl directory
Pull Request -
State: closed - Opened by PeterSH6 about 2 months ago
#44 - [doc] add a new quickstart section
Pull Request -
State: closed - Opened by PeterSH6 about 2 months ago
#43 - [example] add a split placement tutorial
Pull Request -
State: closed - Opened by PeterSH6 about 2 months ago
- 1 comment
#42 - Are optimizer states reloaded or offloaded during the conversion from actor training to actor rollout?
Issue -
State: closed - Opened by G1aZzz about 2 months ago
- 1 comment